Estuary salinity prediction using a Support Vector Machine based approach: a case study of the Po di Goro estuary

The salt-water intrusion in the river estuaries can damage freshwater ecosystems and compromise the many anthropogenic activities which rely on the exploitation of the delta's rivers. Moreover, estuaries' salinization is expected to worsen due to climate change. Thus, a tool able to predict the estuaries' salinization would be essential. In the present study, to face this problem, we propose a machine learning approach, based on Support Vector Machine (SVM), to predict the estuaries' salinity, starting from a set of input variables. Models developed used four-year salinity observations at the Po di Goro estuary (Po River, Italy) and performances are compared to the predictions obtained by the physics-based model CMCC EBM. The performance reached with the machine learning approach is remarkable, with a R-2 of 0.79 and root mean square error of 3.33 psu, performing better than the physics-based model in terms of prediction performance. Since the developed models can accurately perform the estuaries' salinity predictions, we believe that the proposed strategy can be successfully applied to forecast the estuary salinity at different river mouths.